effects of stress and balance of options on decision...
TRANSCRIPT
Department of Physics, Chemistry and Biology
Master Thesis
Effects of stress and balance of options on
decision-making and associated physiological
responses in laying hens
Mia Persson
LiTH-IFM-A-Ex--12/2634--SE
Supervisor: Christine Nicol, Bristol University; Per Jensen,
Linköping Univesrity
Examiner: Dominic Wright, Linköping University
Department of Physics, Chemistry and Biology
Linköpings universitet
SE-581 83 Linköping, Sweden
Rapporttyp
Report category
Examensarbete
D-uppsats
Språk/Language
Engelska/English
Titel/Title:
Effects of stress and balance of options on decision-making and associated physiological responses in
laying hens
Författare/Author:
Mia Persson
Sammanfattning/Abstract:
Animal preferences in choice tests have frequently been used within animal welfare research to make
recommendations about animal handling and husbandry. It is therefore important that these results are
obtained in a way that, as far as possible, respects the behavioural capabilities of the animal. Stress has
been shown to affect cognitive processes in animals and could therefore affect the decision making
process. To examine the effects of stress on decision making, 16 laying hens were trained to distinguish
between two different quantities of a food reward. A decision balance point was found, by increasing the
cost of reaching the large reward, in lines with the theory of demand curves. Hens were then tested in a t-
maze choice test with both balanced and unbalanced sets of options, with and without prior stress
treatment. Choice, latency to choose, heart rate and temperatures were recorded. Hens that received stress
treatment prior to their first test session were affected by this even in subsequent sessions where they did
not receive stress treatment. This effect was not found in hens that first received stress treatment prior to
their second test session. This shows the influence of previous experiences on animal decision making.
Also, a decrease in heart rate during the decision making period was found, when making a choice
between balanced options, indicating anticipation of difficulty. Additionally, this shows that physiological
measurements such as heart rate could be of importance for future studies and greater understanding of
underlying processes of animal decision making.
ISBN
LITH-IFM-A-EX—12/2634—SE
__________________________________________________
ISRN
__________________________________________________
Serietitel och serienummer ISSN
Title of series, numbering
Handledare/Supervisor Per Jensen
Ort/Location: Linköping
Nyckelord/Keyword:
Choice test, decision making, heart rate, laying hen, push-door, stress, t-maze
Datum/Date
2012-05-25
URL för elektronisk version
Institutionen för fysik, kemi och biologi
Department of Physics, Chemistry and
Biology
Avdelningen för biologi
Instutitionen för fysik och mätteknik
Contents 1 Abstract ................................................................................................... 1
2 Introduction ............................................................................................. 1
3 Materials and Methods ............................................................................ 6
3.1 Animals, Housing and Husbandry ..................................................... 6
3.2 Habituation, training and defining individual decision difficulties ..... 7
3.3 Experimental Setup ........................................................................... 8
3.3.1 T-maze choice test ....................................................................... 9
3.3.2 Stress Treatment ........................................................................ 10
3.4 Data Collection................................................................................ 11
3.4.1 Choice ....................................................................................... 11
3.4.2 Latency...................................................................................... 11
3.4.3 Heart Rate Monitoring ............................................................... 11
3.4.4 Thermal Imaging ....................................................................... 13
3.5 Ethical Note .................................................................................... 13
3.6 Statistical Methods .......................................................................... 13
4 Results................................................................................................... 13
4.1 Choice ............................................................................................. 14
4.2 Latency............................................................................................ 16
4.3 Heart Rate ....................................................................................... 17
4.4 Thermal Images ............................................................................... 18
5 Discussion ............................................................................................. 18
6 Acknowledgements ............................................................................... 22
7 References ............................................................................................. 22
Appendix 1 ............................................................................................... 25
Appendix 2 ............................................................................................... 27
Appendix 3 ............................................................................................... 28
1
1 Abstract
Animal preferences in choice tests have frequently been used within animal
welfare research to make recommendations about animal handling and
husbandry. It is therefore important that these results are obtained in a way that,
as far as possible, respects the behavioural capabilities of the animal. Stress has
been shown to affect cognitive processes and could therefore affect the decision
making process. To examine the effects of stress on decision making, 16 laying
hens were trained to distinguish between two different quantities of a food
reward. A decision balance point was found, by increasing the cost of reaching
the large reward, in lines with the theory of demand curves. Hens were then
tested in a t-maze choice test with both balanced and unbalanced sets of options,
with and without prior stress treatment. Choice, latency to choose, heart rate and
temperatures were recorded. Hens that received stress treatment prior to their
first test session were affected by this even in subsequent sessions where they
did not receive stress treatment. This effect was not found in hens that first
received stress treatment prior to their second test session. This shows the
influence of previous experiences on animal decision making. Also, a decrease
in heart rate during the decision making period was found, when making a
choice between balanced options, indicating anticipation of difficulty.
Additionally, this shows that physiological measurements such as heart rate
could be of importance for future studies and a greater understanding of the
underlying processes of animal decision making.
1.1 Keywords Choice test, decision making, heart rate, laying hen, push-door, stress, t-maze.
2 Introduction
Decision making in animals has been described as a process whereby an animal
chooses one specific behavioural response from a set of potential alternatives
(Dill, 1987). Animals make decisions on a daily basis in a wide range of
contexts including when and where to feed, sleep or court or when to engage in
activities which could potentially expose them to a predation risk (McFarland,
1977; Martel and Boivin, 2011). These decisions are very likely to influence the
animals’ life span and fitness (Blumstein and Bouskila, 1996).
Understanding the process of animal decision making is invaluable in
behavioural ecology, to understand and in some cases even predict how animals
may respond to environmental changes (Blumstein and Bouskila, 1996).
Behavioural ecology models describe how an animal should choose to
maximally utilize e.g. energy resources (McNamara and Houston, 1987).
However, when studying animal decision making, inconsistencies can be seen
where animals do not always make the rational decisions predicted by the
model. This occurs even though the animal is considered to be able to evaluate
2
trade-offs and make a rational decision (Blumstein and Bouskila, 1996; Schuck-
Paim et al. 2004). It is important to take in consideration that animals’
motivational priorities are not fixed and can therefore result in behavioural
flexibility shown as inconsistency in their choices (Bateson, 2004; Browne et al.,
2010). Motivational priorities and preferences may vary with species, breeds and
sexes. Even within the same sex of the same species, preferences can change
with the time of the year, experiences, age and reproductive state. Also, other
factors like energy and time budgets, and if the animal is stressed or not, can
affect motivational priorities, especially if the costs associated with acquiring a
resource is altered e.g. the positioning of the cost or if the length of time
accessing the resource is increased (Warburton and Nicol, 1998). Availability of
cues (e.g. visual, tactile, odours) can also have an effect on preferences in
animal choice tests. In other words, there are many factors that can affect
animals’ choices and therefore need to be taken into account when studying
animal decision making.
To better understand how fluctuation in motivational priorities can affect
decision making, a model like the one presented by Blumstein and Bouskila
(1996) can be used. They use the term assessment, defined as the process when
animals evaluate perceived stimuli and convert these into what they call an
informational state. In other words, assessment is the process of acquiring the
information. What is important to remember is that assessment results in an
informational state. Decision making is what follows the process of assessment.
An animal evaluates the informational state according to its current state as
explained earlier (age, experience, etc.) and follows up with a behavioural
response that we see as the decision. The process of decision-making involves
evaluating trade-offs where the animal weighs benefits and costs against each
other to, possibly, make a rational decision which we can record as the
behavioural response. When two animals react with different behavioural
responses to the same stimuli, the model presents us with two different possible
explanations. Either the animals are in two different states and therefore assess
the stimuli in two different ways and end up at different informational states
from the same stimuli. The second explanation is that the animals perceive the
stimuli in a similar way and end up with the same informational state but for
some other reason make different decisions, resulting in two different
behavioural responses. So far there is no way to indicate which one of these two
explanations is more or less relevant because there is no method for empirically
measuring the informational state of an animal. Instead, animals have been
treated as a “black box” where we put in a stimulus and the outcome is a
behavioural response. However, these processes of assessment of informational
states underlying decision making in animals may be the key mechanisms that
we need to understand and predict behavioural responses. It has, however, been
discussed whether physiological measurements may be the way to evaluate the
3
process of assessment and the informational state in animals (Bechara and
Damasio, 2005) but so far no studies have been able to show this.
It is believed that choices made by domestic animals are affected by the
individuals’ subjective experiences of suffering or pleasure (Dawkins, 1990).
Therefore, animal decision making can be used to gather important information
about animal welfare (Browne et al., 2010). Choice tests are commonly used in
animal welfare research for evaluating animal preferences (Bateson, 2004).
Choice tests can involve two or more options between e.g. different
environmental conditions, food items or the access to perform certain
behaviours. An option is considered to be preferred by the animal if it chooses it
more often, spends more time with it or approaches it with a shorter latency than
the other options (Brown et al., 2011). The strength of the behavioural
preference for a specific option is thought to show the strength of the animals’
motivation. When an animal cannot perform a behaviour that it is motivated to
do, suffering may occur, affecting animal welfare. Results from preference tests
can be used to make recommendations about animal handling and husbandry
(Bateson, 2004; Fraser and Nicol, 2011). It is therefore essential that these
results are interpreted in a way that, as accurately as possible, reflects the value
of the option for the animal from a welfare point of view.
Since choice tasks and preference tests are popular methods of evaluating
animal preferences and welfare it is important to know which factors that may
influence the outcome of these tests. The disruptive effect of stressors on
cognitive processes, such as decision-making, is an important research area as
this may lead to husbandry and welfare problems (Mendl, 1999). If housing and
husbandry cause stress to laboratory, farmed or zoo animals, this may impair
their cognitive capabilities such as memory and learning. Stress has previously
been shown to affect several cognitive functions where the effect on learning
and memory are the most studied topics in both animals and humans (Graham,
Yoon and Kim, 2010; McEwen and Sapolsky, 1995). A commonly encountered
explanatory model of the link between stress and cognition is the Yerkes-
Dodson law (Mendl, 1999). The Yerkes-Dodson law explains the general
relationship between an individual’s state of arousal or stress and cognitive task
performance as an inverted U-shaped curve. According to this law, a certain
amount of stress or arousal could be helpful in solving a cognitive task while too
high levels of stress or arousal could have the opposite effect. This biphasic
effect is thought to be partly triggered by the two different types of receptors for
the glucocorticoid corticosterone produced as a part of the stress response
(Gross, Siegel and DuBose, 1980; McEwen and Sapolsky, 1995). Type I
receptors have high affinity and react to the everyday fluctuations of plasma
corticosterone while the Type II receptors with lower affinity only react to stress
4
levels. The corticosterone peak in laying hens, after being exposed to a stressor,
has been shown to occur after 15 minutes (Fraisse and Cockrem, 2006).
There is no definite definition of stress and different scientists use different
explanations (Moberg, 2000; von Borell, 2001). Some authors distinguish
between good or bad stress, some focus on mental stress while others use a
broader definition saying that stress is the activation of the Hypothalamic
Pituitary Adrenal – axis (HPA-axis). I will use the latter, broader, definition of
stress due to the difficulty of segregating stress and arousal since the
physiological response is very similar. Therefore, a stressor is defined as
something causing the activation of the HPA-axis in an animal. In addition to
the problem of defining stress it can also be difficult to measure since many of
the physiological stress responses that are recorded also occur in other situations
(von Borell, 2001). For example, an increase in heart rate could be the
autonomic stress response of the cardiovascular system or it could be the result
of an increase in metabolic rate due to e.g. exercise (Broom and Johnson, 1993).
When studying the stress response it is therefore important to use several
measurements.
One way in which stress can affect decision-making in animals is by its effect on
the animals’ attention (Mendl, 1999). Stress has been shown to cause attention
narrowing, shifts and lapses e.g. in a detour test in chicks, random responses are
observed due to stress caused by social isolation (Regolin et al., 1995). A
possible explanation for this random behaviour is that the stress causes a shift in
attention from the task which results in poor performance. Attention narrowing
or shifting can lead to decision-making based on inadequate information
(Mendl, 1999). This is thought to be due to the state of stress or arousal causing
a speed-accuracy trade off, where an increase in response speed results in a
decision being made before all relevant information has been processed.
Another possible effect of stress on decision making is that it may cause a shift
in the animals’ motivational priorities which could result in it preferring another
option than if it were not stressed. Thus far, few studies have focused on the
effect of stress on decision-making in animals and most studies involving choice
tests use the “black box” concept, only studying behaviour and not physiology.
To investigate the effects of stress on decision-making we can use an
experimental model consisting of two different types of decisions that here will
be referred to as “easy” and “difficult” decisions. Easy and difficult decisions
can be defined in several different ways. For example, a difficult decision can be
one which involves a critical outcome, requires processing a large amount of
information where the options are finely balanced. For the present study a
difficult decision was defined as a choice where the options were finely
balanced and, correspondingly, an easy decision where the options were
5
unbalanced (Figure 1). As long as the two resources being compared satisfy the
same need, the substitutability can be evaluated by measuring the cross-point
between the two demand curves (Sørensen et al. 2004). This study was carried
out alongside another which aimed to develop a method for defining decision
difficulty by finding the empirical cross point between two demand curves in
laying hens.
In addition to this, physiological responses were studied by measuring heart rate
(HR) and temperatures. An increase in heart rate (HR) can be an autonomic
response to stress (Moberg, 2000). HR can increase when the level of physical
activity increases or when the metabolic rate increases (von Borell, 2001).
Incidentally, HR can also increase or decrease before the action occurs as an
emotional response where the sympathetic nervous system causes an increase
(tachycardia) and the parasympathetic nervous system causes a decrease
(bradycardia) (Broom and Johnson, 1993). Additionally, a depression in
behaviour can be seen when animals fail to cope with a stressor which can be
accompanied with a decrease in HR (Henry, 1993). Furthermore, core body
temperature can increase as a result of stress, while peripheral body temperature
decreases (Broom and Johnson, 1993). This response is called stress-induced
hypothermia (SIH) and is a result of peripheral vasoconstriction caused by the
sympathetic nervous system. SIH has been shown to occur in many different
species of animals as well as humans and is caused by a wide variety of
stressors, both physical and psychological (Bouwknecht et al., 2007).
Figure 1. Predicted demand curve showing how often an option is chosen with increasing cost to reach Option 1. A and C are unbalanced options, defined as an easy decision. B is the point where the options are finely balanced, defined as
a difficult decision.
6
The aim of this study is to investigate the effects of stress on decision-making in
laying hens. By looking at both physiological and behavioural responses during
both “easy” and “difficult” decisions, this may result in a deeper understanding
of the underlying processes of decision-making.
3 Materials and Methods
3.1 Animals, Housing and Husbandry Sixteen Columbian Blacktail layers were obtained at point of lay (16-18 weeks
of age). On arrival hens were treated with preventative red mite powder,
weighed and given leg tags for identification. Hens were housed in groups of
four and all four groups were housed in the same room in individual pens (1.02
x 1.25 m and 2 m high). The home room was set up as shown in Figure 2, there
were four pens on each side and the hens were held in every other pen so that
the opposite pen was empty. During weekly cleaning the hens were switched
over, to the opposite pen, to avoid housing side bias. One hen was housed
separately, in the opposite pen to her original housing pen, after four weeks due
to injurious pecking behaviour.
Figure 2. Home room (left) with pens and group habituation tunnel set up and procedure room (right) with reward pens, t-maze and logging equipment.
The hens were kept at a light schedule of 12 L : 12 D (light period between 7 am
and 7 pm) in a temperature of 20 ± 2 ˚C. Feed (Farmgate Layers Mash, BOCM
Pauls, Ipswich, Suffolk, UK) was provided ad libitum through two separate
external feeders (0.5 m respectively 0.27 m in length) reached through an
opening at the height of 0.15 m from the pen floor. Water was provided from a
hanging drinker at height 0.2 m from the pen floor. The pens were also provided
with a round perch at height 0.25 m that stretched through the width of the pen,
and a cardboard nest box (0.39 x 0.38 x 0.47 m). In addition, the hens were
provided with a bedding of wood shavings (approximately 10 cm depth).
7
3.2 Habituation, training and defining individual decision difficulties After a settling in period of five days, the hens were gradually habituated to the
different elements of the test procedure. Habituation criteria were set (see
Appendix 1) which had to be fulfilled before progressing to the next stage. This
was assessed on an individual basis. The total habituation time ranged from four
to five weeks before the training phase began. All hens were ready for testing
after eight to nine weeks of training (including push door calibration) and could
then be tested to evaluate the individual decision difficulties.
The theory of demand curves was used to define unbalanced and balanced
options for food acquisition (Sørensen et al. 2004). The balanced options were
estimated by finding the point where two different quantities of the same food
reward became substitutes for one another by altering the cost of reaching the
larger quantity (Figure 1).
The training, calibration and testing phases, described in this paragraph, were
carried out as parts of another experiment which aimed to individually identify
easy and difficult decisions. The hens were individually trained to distinguish
between two different quantities of sweetcorn presented in different coloured,
different sized bowls (Figure 3). The colours and sizes of the bowls were
counterbalanced so that four hens had each of the different combinations. The
large quantity consisted of six pieces of sweetcorn while the small quantity
represented only one piece. When the hens had successfully completed the
training phase and were considered able to distinguish between the two
different quantities of food, they moved on to the calibration phase. Hens were
first calibrated for the "easy" decision (unbalanced options, un-weighted push
doors). Since there were no additional costs to obtain the larger reward hens
were predicted to prefer this option and the calibration threshold were therefore
set to choosing the large quantity 90-100% over at least 12 consecutive trials as
some divergent choices can be expected. When this criterion had been reached,
the calibration phase for the "difficult" decision began (balanced options,
weighted push door). This involved gradually increasing the weight on the push
door to the large quantity. The weight of the push door was adjusted until hens
began to switch their preference from the large quantity to the small quantity. A
balance point was defined as the point which hens chose the large quantity 25-
75 % of the time. Hen 8 however, fell outside of this range and chose the large
quantity 16.7 % of the time but was still tested and included in the analysis since
she performed within the range during testing. After calibration of both the
balanced ("difficult") and unbalanced ("easy") options, the hens were tested
(separate experiment) to confirm the cross-over point of the demand curves. The
push door weights obtained here were then used in the present experiment for all
individuals with the exception of hen 6 who did not push through any of the
8
weighted doors during the testing period and were therefore down regulated to a
slightly lower weight previously used during training.
Figure 3. The different set ups of food bowl size, colour and sweetcorn quantities (A-D).
3.3 Experimental Setup A test was designed to investigate the effects of stress on decision-making in
laying hens. Every individual was tested during both "easy" and "difficult"
choices (unbalanced and balanced) with and without stress treatment (Table 1).
Every hen had four test sessions, one for each combination (control treatment,
easy choice; control treatment, difficult choice; stress treatment, easy choice;
stress treatment, difficult choice). The order of the combinations were
systematically alternated and counterbalanced.
Tabel 1. Option balance and stress treatment combinations.
Option Balances Treatments Combinations
Balanced X Stress
Balanced Stress Balanced X No Stress
Unbalanced No Stress Unbalanced X Stress
Unbalanced X No Stress
The same animals and experimental setup were used in both the current study
and the previous experiment when individual decision difficulty was defined.
Choice tests were carried out using a transparent t-maze tunnel with built in
push doors on each side of the maze, regulated by externally powered
electromagnets (Figure 4). At the end of each arm of the t-maze, the tunnel exits
into a pen, of the same type used for housing, containing shavings and a drinker.
The food bowls containing the reward was placed in the pens, one on each side,
9
25 cm from the entrance and 10 cm from the side wall, clearly visible from
inside the start box.
Figure 4. Experimental setup: the start box with closed side doors and tunnel slide door, t-maze with two transparent sides, push doors and the openings into the pens where the rewards were placed on both sides.
3.3.1 T-maze choice test
The t-maze is commonly used in preference tests where the animal gets to
choose between two options. The animal is placed in the middle of the maze and
can then choose to go either down the right or the left arm, of the maze, to reach
the options. Two experimenters were required to perform the test procedure. The
hen was placed in a start box with sliding doors on the sides facing the openings
of the pens allowing the hen see the reward on each side. The side doors were
closed when the hen was placed in the start box and two further screen doors
were placed either side of the start box to occlude the hens view on entering the
start box. When Experimenter 1 started the HR monitor logging Experimenter 2
simultaneously started the stopwatch and was then responsible for time
recording throughout the test. The actual test started after the thermal imaging
camera was focused, at the onset of thermal imaging recording (approximately
15 seconds after the hen entered the start box) (Figure 5). Ten seconds were
given to record baseline heart rate and thermal images before the first side door
was removed, allowing the hen to see the reward on one side. After five
seconds, the side door was replaced and the second side door was removed,
10
allowing the hen to see the reward on the other side of the maze. When the
second side door had been open for five seconds, both side doors were removed,
allowing the hen to see both options. When the hen had been presented with
both options for the duration of ten seconds the tunnel slide door was opened,
allowing them to enter into the t-maze tunnel. The order of the side door
removal as well as the reward side placement was systematically alternated and
counterbalanced. If the hen had not entered the tunnel within 60 seconds she was
gently pushed inside and the tunnel door were closed. The hen was allowed a
maximum of four minutes in the tunnel. If a decision had not been made within
the duration, the hen was gently removed through the start box and were then
handled in the same way as if removed from the pen. Upon reaching either of
the pens, the hen was given 90 seconds to consume the reward and to regain
baseline heart rate before tested again. Each of the four sessions for each hen
consisted of: two forced tests, one to each side (for the hens to experience the
weight of the push doors), followed by five consecutive free choices before
switching the reward sides and repeating the same procedure. In total, each of
the four test sessions had 14 tests (four forced and ten free). The day before the
testing period started, the hens were each given four free choices as a reminder
of the test procedure used in the previous experiment.
Figure 5. Time line describing the test procedure from the start of HR logging until the tunnel door opening specifying the role for both experimenters.
Before each test session, the hens were food deprived according to a schedule
obtained during the previous experiment (see Appendix 2). It was determined
during the previous experiment that different individuals displayed different
motivational levels to perform the test, therefore, the optimal food deprivation
time was established on an individual basis. The duration of food deprivation
varied between one hour twenty minutes and six hours thirty minutes.
3.3.2 Stress Treatment
The hens were stressed using a form of physical restraint where the hen was
placed in a free hanging net for 5 minutes directly before testing. The hen was
removed from the net and directly brought into the procedure room where a
harness with heart rate monitor was fitted and the test started. This stress method
has previously been shown to cause a significant increase in plasma
11
corticosterone in chickens (Karlsson et al., 2011). Stress treatment was carried
out in a separate room, opposite to the procedure room, to avoid disturbing other
hens.
3.4 Data Collection
3.4.1 Choice
Two different choices were recorded for each test. The first choice which
represents the side of the first push on a push door and the final choice which is
recorded as the side of the pen the hen actually entered.
3.4.2 Latency The latency for an animal to approach an option is a well known behavioural
measure of motivation used in preference tests (Dawkins, 1990; Bateson, 2004).
If an animal approaches an option with a shorter latency than the other, it can be
interpreted as if the animal prefers this option. Also, latency to choose has been
shown to be positively correlated with consistency of choice in hens (Browne et
al., 2010). Two latencies were measured in this study; the latency from the
opening of the tunnel door until the hen first pushed on a door and the latency
until the hen entered a pen.
3.4.3 Heart Rate Monitoring Heart rate (HR) was monitored using the telemetric logging system presented by
Lowe et al, 2007. A Lycra harness, with an integrated pouch, was used to attach
the heart rate monitor to the hens (Figure 6). The harness was specifically
designed for this purpose so as not to restrict the crop or the hens’ movement.
The base unit, which was attached to the computer received the radio signal
from the logging unit and was placed next to the start box for the duration of
each test.
12
Figure 6. One of the hens wearing a harness with a heart rate monitor, standing beside the base unit connected to a computer logging the heart rate.
ECG pads were placed on each of the
featherless patches on both sides of the
keel bone (Figure 7). When needed,
downy feathers were removed in
accordance with home office licence
procedures. The patches were wiped
with Surgical Spirits to remove dead
cells to ensure good contact between
the ECG pad and the surface of the
skin. RVC Telemetry Logger Control
version 1.5 was used for logging the
heart rate. These files were later
analysed using Spike 2 6.10. Three
different ten seconds heart rate
intervals were analysed. The first
interval was basal heart rate when the
hen was in the start box but had not
yet been presented with the options.
The second interval was the decision
Figure 7. Placement of the ECG pads (blue) on either sides of the keel bone.
13
period when the hen could see both options at the same time before allowed to
enter the tunnel. The third interval analysed was the first ten seconds after the
hen entered the chosen pen.
3.4.4 Thermal Imaging
In the current study, stress-induced hypothermia was studied with a non-
invasive method using a thermal imaging camera FLIR SC305, FLIR System
AB, Sweden. The camera was placed as shown in Figure 1 and the back of the
start box was made of black plastic (emissivity 0.292) allowing thermal imaging
to take place through the plastic so the hens could not see the camera. Maximum
head and comb temperatures as well as eye temperatures were analyzed using
FLIR R&D ResearchIR. Basal temperatures were recorded for the same ten
seconds interval as the basal heart rate (in the start box before presented with the
options). The second ten seconds interval was recorded during the decision
phase when the hen was presented with both options at the same time before
allowed to enter the tunnel. Thermal images were recorded with nine frames per
second. From each ten second interval, one good image with a side shot of the
head in focus was chosen for analysing.
3.5 Ethical Note All work was conducted under UK Home office licence (PPL number: 30/2779).
3.6 Statistical Methods
The statistical analysis was carried out using the statistical package IBM SPSS
Statistics 19. Percentage differences between basal values and decision phase
and pen heart rate (HR) were calculated. However, mean HR values were used
when studying the variation of HR over the ten consecutive free choices.
Individual slope values were calculated and analysed using univariate general
linear model. Percentage differences between basal and decision phase values
were calculated for head, eye and comb temperatures. One-way repeated
measurements ANOVA was used to explore and analyse the obtained data.
Preliminary analysis were made looking at possible effects of session order, side
placement and comparing the first five free choices and the second five free
choices. If no significant effects of these factors were found they were excluded
from the statistical models used to examine the effects of stress and decision
difficulty.
4 Results All data was normally distributed. Factors were addressed individually before
taken into the model. No significant differences were found between the side
placements of the rewards and there were no differences between the first five
choices and the second five choices within each session. Also, no differences
were found between the individual test sessions.
14
4.1 Choice When looking at towards which side the first push were made during each test
(ten test per session, 40 tests in total), on average 78.7±4.4(SE) % of these were
towards the large quantity reward. Furthermore, when looking at the pen entered
during each test, on average 76.3±4.5(SE) % of these were large quantity reward
choices.
Two hens did not make all of their choices after receiving stress treatment in
combination with the difficult decision. Hen 2 made six of ten first pushes and
entered a pen five out of ten times with stress treatment and the difficult
decision. During her next session where she did not receive stress treatment but
still had a difficult decision she made eight of ten complete choices. Hen 10
made five of ten first pushes and entered a pen three out of ten times after
receiving stress treatment in combination with the difficult decision. On her
upcoming session where she did not receive stress treatment but had the difficult
decision she did not make any choices at all. On her second upcoming session,
hen 10 had the easy decision and no stress treatment and she made eight out of
ten complete choices.
The stress treatment did not have any significant effect on either the latency to
the first push or until the pen was entered. On the other hand, the percentage of
large quantity choices was significantly higher when the decision was between
unbalanced options (easy decision) than when the decision was between
balanced options (difficult decision). This applies both for the latency until the
first push and the entering of the pen (Figure 8) (first push: F(1)= 5.931; P<0.05;
pen entered: F(1)=19.794; P<0.05).
Figure 8. Difference in direction of the first push (F(1)= 5.931; P<0.05) and pen entered (F(1)=19.794; P<0.05) depending on decision difficulty. Error bars displays standard error.
15
Statistically significant differences were found between the decision difficulty of
the first test session and towards which option the first push was directed
throughout all four sessions (Figure 9). Hens that had a difficult decision during
the first test session had a higher frequency of choosing the large quantity food
reward as direction of their first push (F(1)=12.215; P<0.05). Correspondingly, a
statistically significant difference was found between the pen entered and the
decision difficulty of the first test session (Figure 9). Hens that had a difficult
decision during their first test session had a higher frequency of choosing the
pen containing the large quantity food reward (F(1)=7.746; P<0.05).
The stress treatment that the hens received immediately before their first test
session had a significant effect on both towards which option the first push were
directed and pen entered throughout the entire experiment. Hens that received
stress treatment immediately before their first session chose the large quantity
reward less frequently (Figure 10), as their first choice, than hens that did not
have stress treatment on their first session (F(1)=7.269; P<0.05).
Correspondingly, hens that received stress treatment immediately before their
first session chose the larger quantity reward less frequently (Figure 10), as their
final choice, than hens that did not have stress treatment on their first session
(F(1)=8.099; P<0.05).
Figure 9. Difference in direction of the first push (F(1)=12.215; P<0.05) and pen entered (F(1)=7.746; P<0.05) depending on whether the hens experienced easy or difficult choices during their first test session. Error bars displays standard error.
16
Figure 10. Differences in first (F(1)=7.269; P<0.05) and final (F(1)=8.099; P<0.05) choices depending on whether the hens received stress treatment or not during their first test session. Error bars display standard error.
4.2 Latency Average latency to the first push was 6.2±1.4(SE) seconds respectively
9.6±1.8(SE) seconds on average latency to enter a pen. The latency until the first
push and until entrance of either of the pens were not affected by the stress
treatment (P>0.1). Latency to first push was not affected by decision difficulty
(P>0.1), however, decision difficulty had a significant effect on the latency to
enter a pen (Figure 11) where it took the hens longer to enter the chosen pen if
the decision was difficult (F(1)=21.282; P<0.05).
Figure 11. Mean latencies (s) to enter a pen for easy and difficult decisions (F(1)=21.282; P<0.05). Error bars display standard error.
Figure 12. Mean latencies (s) to enter a pen depending on whether the hens received stress treatment or not during their first test session (F(1)=7.491; P<0.05). Error bars display standard error.
17
First session decision difficulty and treatment did not have an effect on latency
to first push (P>0.1). First session decision difficulty did not have an effect on
latency to enter pen (P>0.1). However, the stress treatment given immediately
before the first test session had a significant effect on latency to enter pen
(Figure 12) where hens that received stress treatment on their first test session
had longer latency to enter pen (F(1)=7.491; P<0.05).
4.3 Heart Rate
Neither the stress treatment nor the
decision difficulty affected mean
basal heart rate (HR), decision
phase HR or pen HR (P>0.1).
However, there was a strong
quadratic relationship between these
three different HR measures (Figure
13) showing a significant decrease
in mean decision phase HR in
comparison to basal HR and a
significant increase in pen HR in
comparison to basal HR
(F(1)=40.644; P<0.05).
When looking at the overall pattern
of mean heart rate over the ten free
choices there was a negative
gradient slope independent of decision difficulty and treatment (see Appendix 3
Figure 15-18). When looking at the mean decision phase HR over ten free
choices there is a significant linear relationship for both easy decision and
difficult decision HR (Figure 14) (F(1)=4.816; P<0.05). The negative slopes are
significantly different between easy (mean: -1.309) and difficult (mean: -2.389)
choices where the mean HR during the decision phase has a greater decline
during difficult decisions (F(1)=4.815; P<0.05).
Figure 13. Quadratic relationship between mean basal, decision phase and pen heart rate (F(1)=40.644; P<0.05).
18
4.4 Thermal Images
Neither the stress treatment nor the decision difficulty resulted in any
differences in head, eye or comb temperatures which only had a great individual
variation.
5 Discussion To summarise the results, hens that received stress treatment immediately before
their first test session chose the large quantity reward less frequently than hens
that did not receive stress treatment before their first test session. Also, hens that
received stress treatment before their first test session showed a longer latency to
enter either of the chosen pens in comparison to hens that did not receive stress
treatment at their first test session. Decision difficulty, on the other hand, did
have an effect on the overall choices made where the larger quantity reward was
chosen less frequently for difficult choices but still chosen more often than
expected. In addition, latency to enter pen was longer for difficult decisions than
for easy decisions. Also, the decrease in decision phase heart rate (HR) over ten
free choices was greater for difficult choices than easy choices. Hens that
experienced a difficult decision during their first test session chose the larger
quantity reward more frequently than hens that had an easy decision during their
first test session. Looking at how HR varies over single tests, there is a quadratic
relationship between the three different intervals analysed. There is a decrease in
Figure 14. Mean decision phase heart rate over ten free choices for easy and difficult decisions (filled lines). A linear relationship was found for both easy and difficult decisions (F(1)=4.816; P<0.05). There is a statistical significant difference between slopes (dotted lines) for easy (-1.309) and difficult (-2.389) decisions (F(1)=4.815; P<0.05). Error bars display standard error.
19
mean HR from the basal values to decision phase values and a great increase in
mean HR when the hen reached the pen with the reward.
Here, a difficult decision was defined as one which presented two balanced
options. In contrast, an easy decision was defined as one which presented two
unbalanced options. Here, “difficult” decisions resulted in a more balanced
choice distribution than the “easy” decision. This indicates that the difficult
choice had more balanced options than the easy choice. The choice distribution
of the balanced options was within the range used as a criterion during training.
On the other hand, the unbalanced options had a more balanced choice
distribution than what was defined during training but not as balanced as what
was defined as a balanced choice distribution which is very important for the
interpretation of the results from this experiment. It has previously been shown
that hens have the cognitive capacity of making rational choices (Browne et al.
2010). However, hens do not give fixed values to options and so their
preferences can change. Even though we can expect the hens to overall have a
preference for the reward that is the most valuable for them, we cannot expect
every individual to always choose the optimal option (McFarland, 1977).
Preferences may vary with motivational priorities and these are not fixed but can
be altered by e.g. altering the cost to reach a resource (Bateson, 2004). Here we
have caused a situation where the hens have to weigh the trade off between
reward size and the cost of reaching it which has resulted in a more balanced
option set.
Furthermore, hens that were stressed immediately before their first test session
had a more balanced choice distribution, throughout the entire experiment, than
hens that were not stressed before their first session. Stress can impair optimal
behaviour (Graham, 2010), due to its effect on the decision making process
(Mendl, 1999). When an animal is stressed by a stressor that is not only
physiologically stressful but also emotionally stressful it may experience
fearfulness (Fraisse and Cockrem, 2006). Fear as well as stress can alter the
motivational priorities of an animal (Bateson, 2004). A stressed or fearful
animal may prioritise a quick escape path instead of evaluating the options of
reward size. Also, stress can cause shifts in attention compromising animals’
ability to make a decision from the given stimuli which could result in them
choosing e.g. a preferred side instead of a preferred option. In this case, the
stress treatment given during the first test session affected the choices made by
hens throughout the entire experiment. This could indicate that the hens
associate the test with being stressed or fearfulness and this persists all the way
through their four separate test sessions. Even so, most of the hens still made
choices and the effect of prior stress treatment did not have major effects on the
outcome of this choice test. This very importantly points out that that prior
handling should not affect choice tests used in animal welfare context.
20
Additionally, hens that had a difficult decision in their first session ended up
choosing the large quantity reward more frequently than hens that had an easy
decision during their first test session. As previously mentioned, it is possible
that the hens’ expectations of the test are influenced by what they experienced at
their first session. If that is the case, hens that expect a difficult choice may be
more confident in pushing through the weighted doors than hens that expect an
easy choice.
Furthermore, there were no effects of decision difficulty or treatment on latency
to first push. However, latency to enter the chosen pen was longer for difficult
decisions which is most likely a result of the fact that the hens had to push
through a weighted door. Hens that were stressed during their first test session
had a longer latency to enter the chosen pen than hens that were not stressed
during their first test session. In addition, these hens had a longer latency to
enter the chosen pen, even though they had a lower frequency of choosing the
larger quantity reward. Previous studies looking at motivation to reach a food
reward in laying hens, found that when hens are less motivated to feed they took
longer time to push through the weighted push doors (Olsson et al. 2002). This
indicates that hens receiving stress treatment in their first session were less
motivated to push through the push doors than hens who did not receive stress
treatment during their first session. This could be a further indication towards
the hens associating the test procedure with stress or fearfulness if they
experienced stress treatment in their first test session.
Moving on, the quadratic relationship in heart rate (HR) over single tests can
most likely be explained as an effect of handling. Basal HR was measured
before the decision phase HR. Therefore, basal HR is likely to be higher due to
the recent handling of the hen. As the hen remains inactive in the start box a
decrease in HR can be expected as a result of relaxation (Henry, 1993). Pen HR
is expected to be significantly increased in comparison with basal values due to
the physical effort of walking through the maze and pushing through the door
causing an increase in sympathetic activity (Moberg, 2000).
Additionally, mean heart rate (HR) over the ten free choices is represented by a
negative linear relationship. This could indicate that hens get more used to the
choice test and relax as a result of confidence in how to control their situation
(Henry, 1993). Unexpectedly, the decrease in HR over the ten free choices
seems more profound when hens are presented with a balanced or “difficult”
choice. A study on learning in dwarf goats found a decrease in HR as a result of
what they call frustration in the beginning of a learning task before the animals
had learned how to cope with the problem of the task (Langbein et al, 2004).
Likewise, Crone et al. (2004) found an anticipatory decrease in HR, in humans,
21
prior to making risky choices. According to the concept of coping by Henry
(1993), the individual cognitive evaluation of a challenging situation can lead to
an activation of the HPA axis due to an uncontrollable situation. This could lead
to a depression in behaviour accompanied with a decrease in HR. It is believed
that the ability to cope with a stressor is mainly controlled by the sympathetic
nervous system and is therefore generally accompanied by an increase in HR
(Porges, 1995; von Borell 2001). This could possibly indicate that the hens did
not find a coping strategy for the balanced choice thus the slowing of HR in
anticipation of difficulty.
Another explanation for the decrease in decision phase heart rate (HR) could be
that what we here defined as a “difficult” and an “easy” decision is actually not
perceived that way by the hens. A decision between two balanced options may
actually be easier than a decision between two unbalanced options. The
differences in decision phase HR could therefore result from the decision
between two unbalanced options being more stressful than the balanced options.
These findings can be interpreted in line with the decision theory presented by
Blumstein and Bouskila (1996) previously mentioned in the introduction.
Behavioural differences (choice and latency) were seen even though the same
stimuli were presented. This could mean that the hens differ in their assessment
of the stimuli and therefore end up with different informational states and so
make different decisions leading to different behavioural outputs. Another
explanation is that the hens do not differ in their assessment of the stimuli and
end up with the same informational state, but when evaluating the informational
state in comparison to their own state (e.g. physiological state) they differ in
their decision making process which results in different behavioural outputs.
From this study we cannot say if the difference is in the informational state or
the hens’ evaluation of the informational state (decision making process).
However, it could be argued that the decrease in decision phase heart rate,
possibly a sign of anticipation of difficulty, could be an indication of differences
in the process of assessment or decision making in the hens. It is therefore
possible that heart rate measures could be one of the keys to solving the “black
box”.
In conclusion, recent experiences related to the test procedure, can have an
effect on hens behaviour in a choice test. This appears to apply even if the hens
do not have any direct negative experiences of the options. Hens expecting a
difficult choice from their first session generally performed better in subsequent
sessions. Hens experiencing stress treatment immediately before their first
session showed a more balanced choice distribution possibly due to a shift in
motivational priorities. Another theory is that the stress treatment prior to the
first test session affected the hens’ informational state, altering evaluation of the
22
informational state, resulting in more balanced choice distribution. However, the
stress treatment did not have a major effect on the outcome of the choice test
indicating the resilience of animal choice test to mild stress such as prior
handling. Additionally, decrease in heart rate during the decision making period,
of a balanced or decision could indicate anticipation of difficulty in hens. This
physiological measurement could be one of the keys to obtain a more detailed
picture of the “black box” that we often face in animal behaviour and welfare
research. Nevertheless, there is still substantial research to be done, on the
effects of stress on animal decision making, to understand underlying processes.
A possible next step could be to investigate the effect of stress on other forms of
“difficult” decisions.
6 Acknowledgements
I thank Prof. Christine Nicol, Bristol University, and Prof. Per Jensen,
Linköping University, for their involvement and good advice. I also thank PhD
Anna Davies, Bristol University, for all the help and experience.
7 References Bateson M (2004) Mechanisms of decision-making and the interpretation of
choice tests. Animal Welfare 13, 115-120
Bechara A, Damasio AR (2004) The somatic marker hypothesis: A neural
theory of economic decision. Games and Economic Behavior 52, 336-372
Blumstein DT, Bouskila A (1996) Assessment and decision making in animals:
a mechanistic model underlying behavioural flexibility can prevent ambiguity.
Oikos 77, 569-576
Von Borell EH (2001) The biology of stress and its application to livestock
housing and transportation assessment. Journal of Animal Science 79, E260-
E267
Bouwknecht JA, Olivier B, Paylor RE (2007) The stress-induced hypothermia
paradigm as a physiological animal model of enxiety: A review of
pharmacological and genetic studies in the mouse. Neuroscience and Behavioral
Reviews 31, 41-59
Broom DM, Johnson KG (1993) Stress and animal welfare. pp 92-95. Kluwer
Academic Publisher, Dordrecht
Browne WJ, Caplen G, Edgar J, Wilson LR, Nicol CJ (2010) Consistency,
transitivity and inter-relationships between measures of choice in environmental
preference tests with chickens. Behavioural Processes 83, 72-78
23
Browne WJ, Caplen G, Statham P, Nicol CJ (2011) Mild environmental
aversion is detected be a discrete-choice preference testing method but not by a
free-access method. Applied Animal Behaviour Science 134, 152-163
Crone EA, Somsen RJM, Van Beek B, Van der Molen MW (2004) Heart rate
and skin conductance analysis of antecendents and consequences of decision
making. Psychophysiology 41, 531-540
Dawkins MS (1990) From an animal’s point of view: motivation, fitness, and
animal welfare. Behavioral and Brain Science 13, 1-61
Dill LM (1987) Animal decision making and its ecological consequences: the
future of aquatic ecology and behaviour. Canadian Journal of Zoology 65, 803-
811
Fraisse F, Cockrem JF (2006) Corticosterone and fear behaviour in white and
brown caged laying hens. British Poultry Science 47, 110-119
Fraser D, Nicol CJ (2011) Preference and motivation research. In Animal
Welfare, 2nd
edition. CAB International, Wallingford UK
Graham LK, Yoon T, Kim JJ (2010) Stress impairs optimal behavior in a water
foraging choice task in rats. Learning & Memory 17, 1-4
Lowe JC, Abeyesinghe SM, Demmers TGM, Wathes CM, McKeegan
DEF(2007) A novel telemetric logging system for recording physiological
signals in unrestrained animals. Computer and Electronics in Agriculture 57, 74-
79
McEwen BS, Sapolsky RM (1995) Stress and cognitive function. Current
Opinion in Neurobiology 5, 205-216
McFarland DJ (1977) Decision making in animals. Nature 269, 15-20
Mendl M (1999) Performing under pressure: stress and cognitive function.
Applied Animal Behaviour Science 65, 221-244
Moberg GP (2000) Biological responses to stress: Implications for animal
welfare. pp 1-21 in The biology of animal stress. CAB International,
Wallingford UK
24
Olsson EAS, Keeling LJ, McAdie TM (2002) The push-door for measuring
motivation in hens: an adaptation and a critical discussion of the method.
Animal Welfare 11, 1-10
Regolin L, Vallortigara G, Zanforlin M (1995) Detour behaviour in the domestic
chick: searching for a disappearing prey or a disappearing social partner. Animal
Behaviour 50, 203-211
Schuck-Paim C, Pompilio L, Kacelnik A (2004) State-dependent decisions
cause apparent violations of rationality in animal choice. PLoS Biology 2, 2305-
2315
Sørensen DB, Ladewig J, Ersboll AK, Matthews L (2004) Using the cross point
of demand functions to assess animal priorities. Animal Behaviour 68, 949-955
Warbuton HJ, Nicol CJ (1998) Position of operant costs affects visits to
resources by laboratory mice, Mus musculus. Animal Behaviour 55, 1325-1333
25
Appendix 1
Table 1. Rough schedule of the habituation, training, calibration and testing.
Week Summary of each week
1 Habituation
2 Habituation
3 Habituation
4 Habituation + Training
5 Training
6 Training
7 Calibration easy choice
8 Push door calibration
9 Calibration difficult choice
10 Testing Experiment 1 (decision
difficulty)
11 Testing Experiment 1 (decision
difficulty)
12 Christmas Holiday
13 Testing Experiment 2 (stress)
14 Testing Experiment 2 (stress)
26
Table 2. Habituation, training and calibration elements and their criteria as well as the approximate number of days and tests needed to fulfil these.
Elements of
habituation,
training and
calibration
Criterion/criteria to fulfil
Number of
days spent on
each criterion
Number
of tests
needed
per
individual
Sweet corn Eats the sweet corn.
1 to 2
Food bowl
Not obviously distressed by the presence
of the bowls, eats sweet corn from the
bowls
3 to 4
Harness plain
Moves around freely without being
obviously distressed by it. Walks forward
without stopping and pecking at the
harness.
4 to 5, 1 hour
sessions
Harness +
HR monitor
Moves around freely without being
obviously distressed by it. Walks forward
without stopping and pecking at the
harness.
1 to 2 in pen, 3
to 4 with test procedure
Tunnel
(group)
Every individual has gone through the
tunnel into the opposite pen and back at
least once.
3 days, 1-1.5
hours a day
Tunnel
(individual)
Walks freely from the start box through the
tunnel without stopping, showing fearful
behaviour or hesitating.
7 to 8 20
Start box
Able to wait for the required duration
without showing fearful behaviour or
distress when removal of side doors and
tunnel door.
6 with gradually
increasing
duration
23
Push door Pushes through the unweighed door using
a correct technique without hesitation. 8 to 10 30-40
Push door
calibration Pushes through the weighted door using a
correct technique without hesitation. 2 to 3 15-20
Easy choice
calibration
Chooses the larger quantity of sweet corn
100-90% over at least 12 consecutive
trials.
4 to 6, some
needed more
training
30-40
Difficult
choice
calibration
Chooses the larger quantity of sweet corn
75-25% over at least 12 consecutive trials.
9 to 10 50-70
27
Appendix 2 Table 3. Duration of food deprivation for all subjects ranging from 1 hour 20 min to 6 hours 30 min.
Hen Food deprivation Hen Food deprivation
1 2 hours 10 min 9 1 hour 20 min
2 2 hours 10 min 10 4 hours 30 min
3 3 hours 11 5 hours 50 min
4 6 hours 30 min 12 1 hour 40 min
5 1 hour 20 min 13 2 hours 10 min
6 2 hours 10 min 14 4 hours 30 min
7 2 hours 30 min 15 1 hour 20 min
8 3 hours 16 2 hours 30 min
28
Appendix 3
Figure 15. Mean basal heart rate for stressed and non stressed hens over ten free choices (P>0.1). Error bars display standard error.
Figure 16. Mean decision phase heart rate for stressed and non stressed hens over ten free choices (P>0.1). Error bars display standard error.
29
Figure 17. Mean pen heart rate for stressed and non stressed hens over ten free choices (P>0.1). Error bars display standard error.
Figure 18. Mean pen heart rate for easy and difficult decisions over ten free choices (P>0.1). Error bars display standard error.